What's Behind The AI Kerfuffle In Publishing?

Sivapriya R is in a fix. She’s given this topic much thought but hasn’t really found a concrete answer. In her role as Publisher, Literary and Translations at Bloomsbury India, she hasn’t felt the need to provide one. So far.

The question is simple: What are the most pressing concerns you face regarding Artificial Intelligence (AI) at work?

“I don't think I've ever received a manuscript that has had AI used in it in any way. It's kind of extraordinary. I don't know, will it change? . . . I haven't actually received a translation done by an AI, so I have no idea how it will look. Any example that I've seen of AI writing seems so transparently unctuous or bad,” she says.

She’s working on a project called 5 Stories, which features short stories from authors across different regional languages like Kannada and Konkani. The Kannada edition has a story by the history-making International Booker Prize winner Banu Mushtaq, the Konkani one has stories by masters like Damodar Mauzo and Pundalik Naik, and she cannot ever imagine them having anything to do with AI.

Even in non-fiction, the authors she works with are people like Anjum Hassan and Yuvan Aves whose craft, research and narrative style she cannot fathom AI ever replicating. And even in fiction, she brings up the deeply human Sakina’s Kiss, Vivek Shanbag’s celebrated story about a man who cheats his wife's younger brother of five acres of fertile land, and how the consequences of that thievery play out in the lives of his son and his granddaughter.

“What's the morality of doing something like that? You can take that land, you can get away with it, but there will be consequences - generation after generation, there will be consequences. I mean, how on earth is AI going to do that? He wrote it in Kannada, but even if he wrote it in English - I mean, is this a story AI can tell? And can it ever tell it like him, with that kind of extraordinary subtlety?” she wonders.

Milee Ashwarya, award-winning Publisher and Group Senior VP of Adult Publishing at Penguin India, who's been pivotal in the country's publishing scene for around 17 years, echoes similar sentiments. “Some of these personal narratives are very difficult to write through AI,” she says, adding: “I think personal narratives need to be written by oneself, and you can tell if it is written by a machine. However, as you've seen in the recent past, I think there have been some cases where books got published…”

She pauses, and I know what she’s referring to. Anybody with even the slightest interest in the publishing industry does.

In March this year, Mia Ballard’s Shy Girl, initially self-published in February 2025 to rave reviews on social media and a 3.52-star rating on Goodreads, was pulled by Hachette (who had acquired the publishing rights) over allegations that AI was heavily used to write the book. Ballard denied AI use, stating that somebody she hired to edit the book had used artificial intelligence. A few months later, the prestigious Commonwealth Short Story Prize was mired in controversy following allegations that as many as three of its award-winning stories were AI-generated. Later, the Commonwealth Foundation stood by its decision and stated that its investigation did not reveal AI use. 

“We have not used AI tools due to concerns regarding artistic ownership and consent surrounding unpublished work from across the Commonwealth. Whilst these tools can serve as useful indicators of potential AI involvement, we also recognise the widely accepted view that they cannot provide conclusive evidence on their own,” said Razmi Farook, the Director-General of the Commonwealth Foundation in a statement.

Last month, Jerry Falade’s debut crime novel Call Me, I’ll Hide The Body was withdrawn from sale by his agents, The Guardian reports, despite a reported $2m offer from Minotaur because they couldn’t wholly substantiate that AI wasn’t used in the writing of the book. Falade has vehemently denied the accusations, stating that “there seems to be a troubling assumption that when a Black writer produces work that attracts significant attention, the work could not possibly be their own.”

In all of this, concerns have been raised from several quarters regarding the validity of AI-detection tools (which are known to be notorious for producing false positives) with an opinion piece by novelist Andrea Bartz in The New York Times - which broke the Shy Girl story in the first place - stating that “we’re barreling towards a rapid erosion of trust between authors and readers” and that as AI tools improve (which they seem to be doing at a meteoric rate), it will become difficult to ascertain what has been written by a human and what has been churned out by an LLM.

Ashwarya also shares these concerns. “I think the challenge with us right now is that AI is also learning so fast. They are also learning the style of a particular writer sometimes, so, therefore, it is definitely tricky. In most cases, the editor can tell. For existing writers, we know their writing because we know their style. We know their weaknesses in language also because we edit them, right? So if those things are not there, you know this looks very different compared to, say, the last three books,” she shares.

At Penguin Random House India, Ashwarya and her team run manuscripts through software to detect plagiarism and AI usage, which she says they use as an indicator and agrees aren’t 100 per cent reliable. If the tools indicate a higher percentage of AI usage, they know that the text warrants a second look. However, even without using such tools, Ashwarya believes one can make out what’s been written by AI. “There is a certain kind of punctuation it throws up, a certain kind of sentence construction. There are some tells for AI writing besides the software that you use. And if you have a doubt, you should always ask the writer. I think it is best to be honest, because these things are very important to be discussed earlier rather than being embarrassed later on,” she adds.

But even before all of this, the agents who approach publishing houses are already aware of the publisher’s policy and do their own checks.

Kanishka Gupta, one of India’s top literary agents and founder of Writer’s Side, the largest literary agency in South Asia, worries that certain high-profile contract cancellations in the US have led to a paranoia of sorts in the publishing industry, especially because even with the best tools, it is not always possible to tell what is AI and what isn’t.

“It's [the rise of AI] just going to create confusion and suspicion amongst agents and editors . . .  what's going to happen is that there may be a clause in publishing contracts - a new kind of an indemnity clause, I think, where authors might be asked to indemnify publishers against use of AI in their writing, and authors may expect the publishers in return to indemnify the author from using AI for editing or for cover design,” he says.

There seems to be general consensus amongst the big-wigs of publishing that using AI to lightly check for grammar or spelling or to do admin tasks like writing emails can be forgiven, given the times we live in. But even then, it is suggested that one avoid AI completely.

“I mean, a good editor is going to edit [the text] himself or herself. You can work on multiple drafts, and they may also contest a lot of edits made by AI. It's not bad for some sort of background checks - fact checking, background check . . . but AI is notorious for making mistakes. I mean, it's just like duplicating work - you get AI to do something, and then you do the same thing,” Gupta adds.

Like Ashwarya, Gupta, too, can make out when a query letter or proposal he gets is written using AI, but some authors, he says, admit to using AI without being asked. “Recently, I got an email from a Marathi writer, saying that his book did very well in Marathi and that he would like me to help him find a publisher for the English translation, which, by the way, had been produced using AI. He was very upfront. I mean, obviously I was not keen on it,” he shares.

AI detection tools are so notoriously unreliable that students in the West have filed lawsuits against their universities after being accused - and won. Sivapriya R speaks about a travelogue she received that had all the markers of AI writing (“the three descriptions one after another, the use of the word "quiet" a number of times, and so on and so forth”), but when she ran it through Bloomsbury’s closed Copilot system, it suggested the text was written by a human.

“This is also something that we need to think about: that there are people who are writing who sound like AI, right? They may not have used it, because LLMs are learning from how humans write. So there's that - the serpent eating its tail kind of thing. It's circular logic,” she says.

Beyond the moral concern of using AI in the writing process and the issue with publishing copyrights (“How can non-original content be fit for copyright? It's not your writing. I think it's a natural burglary,” says Ashwarya), the rise of LLMs is also challenging what it means to be a writer.

“People are saying, ‘Oh, I've come up with the idea, and I'm just using ChatGPT to execute it, big deal’. No, that's a problem, because it's not the same. Even philosophically, it's a very different thing, because if you're using ChatGPT to execute the idea, then ChatGPT is using whatever it's been trained on, so it is effectively plagiarising somebody else's work to produce something ‘new’. An idea in and of itself has limited value. The execution is a very crucial part of how that idea is successfully rendered or not. I think ChatGPT has definitely made a larger number of people feel that the actual writing of something is labour, that it has nothing to do with creativity, it has limited value, it's the idea that is all important, but that's not really the case,” says Ambar Sahil Chatterjee, literary agent and editor at A Suitable Agency who has worked with the likes of Perumal Murugan, Fatima Bhutto and Anees Salim.

Everyone interviewed for this piece is unanimous in agreeing that the best writing comes from human creativity and the human experience - that AI, at least for now, isn’t all that it’s made out to be. 

“I think people should not lose hope in original writing. They should continue to write good books. The human brain is definitely better than a computer brain. Humans created AI, humans created the computer, so let's not underestimate ourselves. I feel very bullish about the future. I think there are great things going to come,” says Ashwarya, who adds that she feels personally invested in this and not just because it’s her job.

Chatterjee recalls a friend telling him that writing comes from a place of anguish, and that’s something AI will never be able to replicate. He doesn’t know what will happen a decade or two from now, but how refined AI becomes in its ability to mimic the human experience will depend entirely on what regulations are imposed on it right now. “Because if it is left unfettered, then who knows?” he wonders.

Regulations are already underway - in accordance with Article 50 of the EU AI Act, companies like Anthropic and OpenAI are now required to watermark AI-generated content, whether that is text, an image or an audio file. Anthropic (the company behind Claude) has already rolled this out with all Claude models launched on or after August 2, 2026, embedding invisible watermarks on any text put into the system, regardless of whether the LLM has written the entire copy or simply added a comma. OpenAI and Google have also rolled out similar systems in compliance with the Act for image and audio files, with text watermarking still in development. It must be noted that the watermarks are applicable to users worldwide, even those not affected by EU laws.

For all the disdain for AI and the belief that the best writing comes from taking it ‘bird by bird’, Gupta says that writers, as an ilk, are intrigued by the possibilities that AI possesses as a fictional character. So much so that he’s sick of the submissions. “Everybody's writing about AI, how AI has impacted this, how AI has impacted that. They are turning these into full-length books. We're going to see a lot of that, where AI captures the imagination of the writer. Even the top writers and editors, while disgusted and repulsed, are endlessly fascinated by AI,” he shares.

He’s not wrong. 

At the very start of our interview, when asked about how AI had impacted her work, Sivapriya R was very clear that what makes a story special is a mix of life experience, the skills cultivated by a writer and their distinct perspective of the world. 

“But I also think - I've thought about this - would I not read a story written by AI? We haven’t really achieved singularity, right? At least not that I know of. At some point, if artificial intelligence, an entity, becomes conscious and tells its own story - that I’m definitely going to be interested in,” she says.